Lipopolysaccharide Transport Involves Long-Range Coupling between Cytoplasmic and Periplasmic Domains of the LptB₂FGC Extractor

Gram-negative bacteria are intrinsically resistant to many antibiotics due to the presence of lipopolysaccharide (LPS) at their cell surface. LPS is transported from its site of synthesis at the inner membrane to the outer membrane by the Lpt machine.

ABSTRACT

The cell surface of the Gram-negative cell envelope contains lipopolysaccharide (LPS) molecules, which form a permeability barrier against hydrophobic antibiotics. The LPS transport (Lpt) machine composed of LptB₂FGCADE forms a proteinaceous transenvelope bridge that allows for the rapid and specific transport of newly synthesized LPS from the inner membrane (IM) to the outer membrane (OM). This transport is powered from the IM by the ATP-binding cassette transporter LptB₂FGC. The ATP-driven cycling between closed- and open-dimer states of the ATPase LptB₂ is coupled to the extraction of LPS by the transmembrane domains LptFG. However, the mechanism by which LPS moves from a substrate-binding cavity formed by LptFG at the IM to the first component of the periplasmic bridge, the periplasmic β-jellyroll domain of LptF, is poorly understood. To better understand how LptB₂FGC functions in Escherichia coli, we searched for suppressors of a defective LptB variant. We found that defects in LptB₂ can be suppressed by both structural modifications to the core oligosaccharide of LPS and changes in various regions of LptFG, including a periplasmic loop in LptF that connects the substrate-binding cavity in LptFG to the periplasmic β-jellyroll domain of LptF. These novel suppressors suggest that interactions between the core oligosaccharide of LPS and periplasmic regions in the transporter influence the rate of LPS extraction by LptB₂FGC. Together, our genetic data reveal a path for bidirectional coupling between LptB₂ and LptFG that extends from the cytoplasm to the entrance to the periplasmic bridge of the transporter.

IMPORTANCE Gram-negative bacteria are intrinsically resistant to many antibiotics due to the presence of lipopolysaccharide (LPS) at their cell surface. LPS is transported from its site of synthesis at the inner membrane to the outer membrane by the Lpt machine. Lpt proteins form a transporter that spans the entire envelope and is thought to function similarly to a Pez candy dispenser. This transenvelope machine is powered by the cytoplasmic LptB ATPase through a poorly understood mechanism. Using genetic analyses in Escherichia coli, we found that LPS transport involves long-ranging bidirectional coupling across cellular compartments between cytoplasmic LptB and periplasmic regions of the Lpt transporter. This knowledge could be exploited in developing antimicrobials that overcome the permeability barrier imposed by LPS.

INTRODUCTION

Gram-negative bacteria have two membranes that serve as permeability barriers against toxic compounds (1). The inner membrane (IM) is a phospholipid bilayer that surrounds the cytoplasm. Outside the IM is the periplasm, an aqueous compartment that contains a thin cell wall composed of peptidoglycan. The periplasm is surrounded by the outmost layer of the cell envelope, the outer membrane (OM). Unlike the IM, the OM is a highly asymmetrical bilayer comprised of an inner leaflet of phospholipids and an outer leaflet of lipopolysaccharide (LPS) (2, 3). LPS is an amphipathic glycolipid that plays a key role in protecting Gram-negative bacteria from hydrophobic antibiotics and detergents (4).

The robust protective qualities of LPS are intrinsic to its structure (4). LPS is composed of three main components, lipid A, the core oligosaccharide, and the O antigen (5, 6). Lipid A is an acylated β-1′-6-linked glucosamine disaccharide that anchors LPS to the membrane. The core oligosaccharide is a set of nonrepeating sugars that is ligated to lipid A. The O antigen is ligated to the outer region of the core oligosaccharide and is a highly variable polysaccharide composed of repeating units of 4 to 6 sugars. Notably, many species of bacteria, including the model organism Escherichia coli K-12 used in this study, do not produce O antigen (7). Both the lipid A and core oligosaccharide portions of LPS are phosphorylated at multiple locations. At the OM, phosphates on neighboring LPS molecules are bridged by divalent cations, facilitating the tight packing of LPS (4). Both the strong lateral interactions between LPS molecules and the amphipathic nature of the glycolipid create a robust permeability barrier at the OM (4).

The synthesis of LPS originates in the cytoplasm (6, 8). The Raetz pathway converts the precursor molecule UDP N-acetylglucosamine to Kdo₂-lipid A (where Kdo₂ represents two molecules of 3-deoxy-d-manno-octulosonic acid) through the action of the Lpx and KdtA enzymes. At the inner leaflet of the IM, Kdo₂-lipid A is then converted into lipid A-core by the Waa enzymes, a series of transglycosylases and kinases that generate the core oligosaccharide. Next, lipid A-core is flipped across the IM by the ATP-binding cassette (ABC) transporter MsbA (9, 10). If O antigen is present, it is independently synthesized in the IM and ligated onto the outer core portion of the lipid A-core molecule in the outer leaflet of the IM by the WaaL ligase (11, 12).

After LPS biosynthesis is complete, the glycolipid is transported from the outer leaflet of the IM to the outer leaflet of the OM by the LPS transport (Lpt) machine, LptB₂FGCADE (Fig. 1A) (13, 14). The Lpt complex spans every compartment of the cell and forms a proteinaceous bridge that allows for the rapid transport of LPS across the cell envelope (15). The IM component LptB₂FGC is an ABC transporter that facilitates the initial extraction of LPS from the IM (Fig. 1B) (16–19). In an ATP-independent manner, an LPS molecule enters a substrate-binding cavity in this transporter. Each round of the ATP-driven cycle is thought to cause the closure of this cavity and the subsequent transfer of an LPS molecule onto the periplasmic bridge composed of the β-jellyroll domains of LptFCAD (15, 16, 20). These β-jellyroll domains are C shaped with a hydrophobic interior that shields the acyl chains of LPS from the aqueous periplasm (21–24). The current model for LPS transport proposes that LPS travels through the periplasmic bridge as a stream. According to this model, when an LPS molecule is extracted from the substrate-binding cavity and placed onto the bridge, it pushes previously extracted LPS molecules in front of it toward the OM (16). Eventually, LPS reaches the OM translocon, LptDE. The lipoprotein LptE dwells within the C-terminal β-barrel domain of LptD, which provides a channel to translocate the sugars of LPS across the OM (23, 25–28).

FIG 1.

LPS is transported to the OM by the LptB₂FGCADE complex. (A) Diagram of the LPS transport (Lpt) machine. ATP binding and hydrolysis by LptB₂ power the extraction and transport of LPS by LptFG. The lipid A portion of LPS is yellow, and the core oligosaccharide portion is orange. (B) Model of the LPS transport cycle by LptB₂FG (depicted as described above for panel A except that, for simplicity, LptC is omitted from this model). First, LPS enters the LptFG cavity. Next, ATP is bound (yellow spheres), causing the concurrent collapse of the LptB dimer and the LptFG cavity. This collapse expels LPS out through the β-jellyroll domain of LptF onto the periplasmic bridge. Finally, ATP is hydrolyzed, which reopens the complex for another round of LPS transport.

The mechanism of LPS extraction by the ABC transporter LptB₂FGC is not well understood. At the most basic level, we know that the cytoplasmic nucleotide-binding domains (LptB₂) bind and hydrolyze ATP, while the transmembrane domains (LptFG) directly interact with LPS via a V-shaped cavity and perform the extraction (Fig. 1B) (29–31). The wall of this LPS-binding cavity is formed by six transmembrane (TM) helices of each LptF and LptG and the single TM helix of LptC. LptC is a bitopic IM protein with a periplasmic β-jellyroll domain that bridges the β-jellyroll domains of LptF and LptA (31–33). The function and location of the TM of LptC throughout the transport cycle are unknown, but it is worth noting that this TM is dispensable for LPS transport (34). Without the TM of LptC present, LPS forms more high-affinity contacts with the LptFG cavity, perhaps priming the transporter for ATP binding (31, 33). As in other ABC transporters, the ATP-driven cycling of the nucleotide-binding domains must be carefully coordinated to the movement of the transmembrane partners in order to facilitate substrate movement (i.e., LPS extraction) and resetting of the transporter (Fig. 1B). These domains are coupled at their interface by direct physical interactions between the groove region of LptB₂ and the coupling helices of LptFG (24, 35). After LPS enters the cavity of LptFG (30–33), LptB₂ binds two ATP molecules and transitions to a closed-dimer conformation (31, 33). This causes the concurrent collapse of the LptFG cavity through a rigid-body style of movement dependent on the physical connections between the LptB groove region and the LptFG coupling helices (31, 33, 36, 37). The cavity closure expels LPS out through the β-jellyroll domain of LptF onto the periplasmic bridge (32). ATP is then hydrolyzed, and ADP and inorganic phosphate are released, resetting and reopening the transporter (Fig. 1B).

Recently, we proposed that the functional coupling between the LptB dimer and the LptFG proteins is bidirectional (37). Not only do interactions between the two LptB monomers and ATP affect the conformational state of the LptFG cavity, but interactions between LPS and the LptFG cavity also affect the conformational state of the LptB dimer. Individual changes to LptB residues E86 and R144, which are located in the groove region, are thought to decrease the ability of the LptB homodimer to close around two molecules of ATP (36, 37). Defects in LPS transport caused by changes to E86 and R144 in LptB can be suppressed by changes to either the structure of lipid A (i.e., removal of acyl chains) or residues in TM helices of LptG that are part of the substrate-binding cavity. These results suggested that interactions between the substrate (LPS) and its binding site in the transporter (LptFG cavity) affect LptB’s function. Furthermore, these changes to the LPS structure and LptG’s TM helices also suppress a defect in the coupling helix of LptF. Consequently, LPS extraction relies on bidirectional coupling mediated by ATP binding by LptB, interactions between the groove region of LptB and the coupling helix of LptF, and interactions between the LptFG cavity and the hydrophobic portion of LPS (37). Here, we further defined regions in LPS and LptFG that are involved in this step of extraction of LPS by LptFG. Surprisingly, we found that changing the structures of the core oligosaccharide of LPS (which does not interact with the substrate-binding cavity) or the periplasmic loop connecting the LptFG cavity to the base of the β-jellyroll of LptF alters the activity of the transporter. We therefore propose that LPS transport involves long-range bidirectional coupling of the ATPase LptB in the cytoplasm to the entrance to the first β-jellyroll of the periplasmic Lpt bridge.

RESULTS

Closure of the LptB dimer is defective in lptB(R144H/G236A) mutants.

We wanted to investigate how the closure of the LptB dimer upon ATP binding is coupled to the closure of the LPS-binding cavity in LptFG by selecting for suppressors that restore LPS transport in lpt mutants that are defective in this step in the transport cycle (Fig. 1B). We can perform such selections by taking advantage of the essentiality of LPS in E. coli and its role in establishing a permeability barrier against hydrophobic antibiotics (4, 36). Cells with fully functional Lpt machinery are highly resistant to hydrophobic antibiotics. If the function of Lpt is partially disrupted, less LPS is transported, sensitizing cells to hydrophobic antibiotics. The degree of antibiotic sensitivity corresponds to the functional state of the transporter. Greatly defective lpt alleles can lead to either conditional lethality, where cells can survive only under slow-growth conditions (38), or cell death under all conditions tested.

We first needed to identify an lpt mutant that we could use in our suppressor selection. Previously, we described how residue R144 in LptB is necessary for ATP binding and closure of the LptB dimer (36). Residue R144 is located in the signature helix of LptB and is part of the signature motif that forms one-half of the ATP-binding site of nucleotide-binding domains of ABC transporters (Fig. 2A and B) (36, 39–42). lptB encoding an R-to-H change at residue 144 [lptB(R144H)] is a conditional loss-of-function allele (Fig. 2C) that results in defects in the binding of ATP and closure of the LptB dimer (36). We also demonstrated that changes to the C-terminal domain (CTD) of LptB suppress the conditional lethality conferred by lptB(R144H) (36). Interestingly, the CTD of LptB is essential for LPS transport. On their own, alterations to the CTD cause defects in the opening of the LptB dimer that occurs after ATP hydrolysis, which manifest as phenotypes that range from severe OM permeability defects (e.g., by adding a C-terminal peptide extension to LptB) to death (e.g., by a G236A change in LptB) (Fig. 2C) (36). Thus, changes to R144 and the CTD of LptB suppress one another because they exert opposing effects on the conformational state of the LptB₂ dimer by favoring either its open-dimer (R144H) or its closed-dimer (CTD alterations) state (36). Combined, the double mutant complexes can transition between the two conformations. However, this cosuppression varies in strength depending on the specific alteration made to the CTD. One of the double mutants, the lptB(R144H/G236A) mutant, still exhibits sensitivity to antibiotics, indicating that LPS transport is not fully restored (Fig. 2C and D). We therefore explored whether the lptB(R144H/G236A) mutant could be used in our suppressor selection.

FIG 2.

lptB(R144H/G236A) is a partial-loss-of-function allele. (A, top) Cartoon representation of the crystal structure of an LptB monomer from E. coli (PDB code 6MBN) bound to ATP (cyan sticks) as viewed from the dimer interface. (Bottom) Side view of same crystal structure showing the LptB dimer with one monomer in gray, with residues R144 and G236 shown as green and red spheres, respectively, and the other monomer in tan, with residues R144H and G236 represented as blue and magenta spheres, respectively. (B) Diagram of the LptB monomer structure from panel A also depicting where the coupling helix of LptFG interacts with LptB. Each ATP molecule binds to a site composed of the signature (Sig) and Walker A (WA) motifs of each monomer. (C) lptB(R144H) and lptB(G236A) suppress one another (originally shown by Simpson et al. [36]). Plasmid-borne lptB alleles were assessed for their ability to complement chromosomal ΔlptB. Partial loss of function (LOF) is observed in haploid strains that grow in LB and minimal media but exhibit OM permeability defects, conditional LOF refers to alleles that cannot complement in LB but can in minimal medium, and total LOF refers to those that do not complement under any condition. (D) The lptB(R144H/G236A) mutant has increased sensitivity to bacitracin (Bac), erythromycin (Ery), and rifampicin (Rif) but is almost as resistant to novobiocin (Nov) as the wild type. OM permeability to antibiotics was assessed via a disc diffusion assay on LB by measuring zones of inhibition surrounding the discs (in millimeters). Zones of complete clearing are represented by the dark bars, and zones of partial clearing are represented by the lighter stacked bars. The dashed line indicates the size of the antibiotic discs (7 mm). Data represent the averages and standard deviations from three independent experiments. The absence of error bars indicates a standard deviation of zero.

To determine whether the LptB^R144H/G236A dimer is still partially defective in closing or opening, we characterized the effect that the antibiotic novobiocin has on the lptB(R144H/G236A) mutant. In addition to targeting and inhibiting DNA gyrase, novobiocin also binds to the groove region of LptB and enhances the rate of LPS transport (36, 37, 43). This effect on LptB allows novobiocin to specifically suppress a subset of lpt mutants that have defects in closing the LptB dimer, including lptB(R144H), but not mutants with defects in the CTD of LptB (36, 37, 43). Since novobiocin is a hydrophobic antibiotic, most lpt-defective mutants show increased sensitivity to this gyrase inhibitor and other hydrophobic antibiotics (34, 35, 44); in contrast, lpt mutants that show near-wild-type resistance specifically to novobiocin, but not to unrelated hydrophobic antibiotics, have a defect in LPS transport that is suppressed by novobiocin (36, 37, 43). As the lptB(R144H/G236A) mutant exhibits this hallmark near-wild-type resistance to novobiocin while still being very sensitive to bacitracin, erythromycin, and rifampicin (Fig. 2D), LptB^R144H/G236A still has defects in achieving the closed-dimer conformation. We thus proceeded to use the lptB(R144H/G236A) mutant in our selection.

The loss of WaaP suppresses defects in LPS transport caused by lptB(R144H/G236A).

To better understand how LptB₂FGC complexes attain the concomitant closure of the LptB dimer and the LptFG cavity to extract LPS, we selected for spontaneous suppressors of the partial-loss-of-function allele lptB(R144H/G236A) by plating on LB agar containing bacitracin (see Fig. S1A in the supplemental material). We identified a suppressor that also showed increased resistance to a variety of hydrophobic antibiotics, indicating rescued functionality of the Lpt machinery (Fig. S1). This mutant has an insertion element located at the 5′ end of waaP (waaP::IS1) that we confirmed to cause a loss of function since a chromosomal ΔwaaP deletion suppresses lptB(R144H/G236A) to the same extent as waaP::IS1 (Fig. S1). We then proceeded to use the chromosomal ΔwaaP allele in the rest of our study to prevent reversion of the waaP::IS1 allele.

WaaP is a kinase that phosphorylates the first heptose of the core oligosaccharide during LPS synthesis, and a deletion of this gene results in an altered LPS structure that lacks several moieties (Fig. S1B) (45). Indeed, disruptions to the core oligosaccharide biosynthesis pathway often result in multiple structural changes because the absence of one moiety can affect some of the downstream reactions catalyzed by different enzymes. For example, the presence of the phosphate on heptose I (added by WaaP) is necessary for WaaQ and WaaY to respectively add heptose III and a phosphate to heptose II, although the rest of the core sugars are still added (Fig. 3A and Fig. S1B) (45). Thus, in waaP strains, LPS lacks heptose III and the phosphorylation of heptoses I and II (Fig. 3B).

FIG 3.

Removal of heptose III is essential for suppression of lptB(R144H/G236A). (A) Diagram of the core oligosaccharide biosynthesis pathway. Proteins examined in this study are color coded corresponding to the portion of the core oligosaccharide that they are responsible for adding (shown in panel B). When waaP, waaQ, or waaY is deleted, the biosynthesis pathway will resume at WaaB. WaaBOJU will add on the last four sugars to LPS despite a deletion of waaPQY in the middle of the biosynthesis pathway. This discontinuous synthesis pattern is represented by the dashed arrow. (B) Table summarizing the relative quality of suppression of lptB(R144H/G236A) by waaGPQY deletion alleles. The genotype (top row), the corresponding core structure (middle row), and the ability to suppress lptB(R144H/G236A) (bottom row) are shown. The relative quality of suppression is indicated by the plus signs. Wild type refers to strain NR4447 [haploid lptB(R144H/G236A) with no waa deletion alleles]. Kd refers to Kdo, H refers to heptose, Gc refers to glucose, and G refers to galactose. (C) ΔwaaQ is the best waa suppressor of lptB(R144H/G236A). A disc diffusion assay on LB was used to characterize the OM permeability of either wild-type lptB (labeled WT) or lptB(R144H/G236A) (labeled +/−) strains carrying waaGPQY deletion alleles as described in the legend to Fig. 2D. Data are the averages from three independent replicates. Error bars represent the standard deviations, and the absence of error bars indicates a standard deviation of zero. NR4446 was used as the parental wild-type lptB strain.

Strains expressing lptB(R144H/G236A) have lower levels of LptB, as previously seen (36). These reduced LptB levels do not significantly change in the presence of the ΔwaaP suppressor (Fig. S2A). Therefore, ΔwaaP does not suppress lptB(R144H/G236A) by altering LptB levels. Changes in total cellular levels of LPS are also not altered in lptB(R144H/G236A) cells by ΔwaaP (Fig. S2B). In addition, we observed that, as previously reported, cells lacking WaaP in an otherwise wild-type strain display increased sensitivity to rifampicin but not bacitracin (Fig. S1) (45–48). This increase in OM permeability caused by ΔwaaP might result from the loss of the charged phosphates on the core of LPS (Fig. S1B) that contribute to the formation of a polyelectrolyte-like barrier against hydrophobic compounds. Together, these analyses failed to reveal an obvious mode of suppression of lptB(R144H/G236A) by ΔwaaP.

The loss of WaaP (ΔwaaP) is also known to activate the regulation of capsule synthesis (Rcs) stress response (49). Therefore, we asked whether the activation of the Rcs response is responsible for the suppression of lptB(R144H/G236A) by ΔwaaP. To test this, we deleted rcsB, which encodes the response regulator of the Rcs stress response (50, 51), and observed no change in OM permeability in the wild-type, the lptB(R144H/G236A) ΔwaaP suppressor, or the single mutant parent strains (Fig. S3). Therefore, the Rcs response does not contribute to the suppression of lptB(R144H/G236A) by ΔwaaP. This leaves us with the simplest explanation that the change in the LPS structure resulting from ΔwaaP is responsible for the suppression of lptB(R144H/G236A).

Changes to the core oligosaccharide structure mediate suppression of defects caused by lptB(R144H/G236A).

Since our data suggested that ΔwaaP suppresses lptB(R144H/G236A) by altering the structure of LPS, we next asked whether this suppression is specific to the loss of WaaP or whether eliminating other Waa enzymes could also suppress this lpt allele. We assessed the OM permeability of strains containing chromosomal deletions of core oligosaccharide biosynthesis genes that function both upstream and downstream of waaP in combination with both wild-type lptB and lptB(R144H/G236A) haploid alleles (Fig. 3). In the presence of wild-type lptB, ΔwaaP and ΔwaaG confer moderate OM permeability defects to several hydrophobic antibiotics, whereas ΔwaaQ and ΔwaaY do not (Fig. 3C and D). As suggested above, the increase in OM permeability with ΔwaaP and ΔwaaG might result from the loss of the charged phosphates in the core of LPS (Fig. 3B) (45, 52). Furthermore, in addition to ΔwaaP, deletions of waaG, waaQ, or waaY also suppress the OM permeability defects conferred by lptB(R144H/G236A) to various degrees. Of these deletions, ΔwaaQ is the best suppressor of lptB(R144H/G236A) and returns the cells to a wild-type phenotype (Fig. 3C and D).

Cells with ΔwaaQ are missing both heptose III and a phosphate on heptose II in the core of LPS (Fig. 3B) (45, 52). Mutants with ΔwaaY produce LPS with heptose III but lack the phosphate on heptose II (45). Since ΔwaaQ fully suppresses defects caused by lptB(R144H/G236A), while ΔwaaY only partially suppresses sensitivity to bacitracin, we concluded that the absence of heptose III is critical for suppression. Indeed, ΔwaaP, ΔwaaG, and ΔwaaQ suppressors all lack this sugar in the LPS core. The partial suppression of bacitracin sensitivity caused by ΔwaaY suggests that the loss of the phosphate on heptose II mildly suppresses and contributes to the robust suppression by ΔwaaQ. However, we cannot test the effect of only missing heptose III, owing to the order of the biosynthesis pathway of the LPS core, as phosphorylation of heptose II requires the presence of heptose III and, therefore, the activities of WaaP and WaaQ (Fig. 3A and B). As ΔwaaQ is the best suppressor of the waa deletion alleles, we used this suppressor going forward in our study.

Suppression by ΔwaaQ is specific to lpt alleles causing defects in the collapse of the LptFG cavity that triggers LPS extraction.

As stated above, lptB(R144H/G236A) encodes both a change (R144H) in the signature helix of LptB that decreases dimer closure and an alteration to the CTD that decreases dimer opening (36). These changes partially compensate for one another, but LptB^R144H/G236A is still somewhat defective in dimer closure because it can still be suppressed by novobiocin (Fig. 2D). Therefore, we hypothesized that ΔwaaQ suppresses defects in LptB dimer closure caused by the R144H, acting in an additive fashion with the alteration to the CTD of LptB.

To test this prediction, we first introduced ΔwaaQ into a single lptB(R144H) mutant. On its own, lptB(R144H) is able to complement the loss of chromosomal lptB only under slow-growing conditions (i.e., a haploid mutant grows only on minimal medium plates) (36). We found that in combination with ΔwaaQ, lptB(R144H) also complements under fast-growing conditions (i.e., on LB plates) (Fig. S4A). Nevertheless, the lptB(R144H) ΔwaaQ double mutant still exhibits strong OM permeability defects, indicating that ΔwaaQ is not a robust suppressor of lptB(R144H) (Fig. S4B). Therefore, full suppression of lptB(R144H) requires both the G236A change to the CTD of LptB and ΔwaaQ, suggesting that these suppressors affect LPS transport by different mechanisms. Indeed, combining ΔwaaQ with a nonlethal defect in the CTD of LptB caused by the addition of a C-terminal His₈ tag preceded by a glutamate (EHis₈) further increases the permeability of the lptB-EHis₈ strain (Fig. S5). Together, these data show that ΔwaaQ is not a general, nonspecific suppressor of Lpt or OM permeability defects but one that partially suppresses defects conferred by the R144H change in LptB.

Thus far, our data suggest that ΔwaaQ affects a specific step in the transport cycle, when the LptB monomers dimerize upon binding ATP, which results in the concomitant closure of the LptFG cavity and the translocation of LPS from the cavity to the Lpt bridge. To thoroughly assess if ΔwaaQ specifically affects this step, we combined ΔwaaQ with lpt alleles that cause defects in various steps in the LPS transport cycle. Variants of residues R144 and E86 in LptB and of residue E84 in the coupling helix of LptF cause defects in collapsing the LptFG cavity (37). We observed that ΔwaaQ fully suppresses the partial-loss-of-function alleles lptB(E86Q) and lptF(E84A) (Fig. 4A to C). However, ΔwaaQ does not suppress the total-loss-of-function lptB(E86A) allele (Fig. 4A), demonstrating that ΔwaaQ is able to suppress only mild to moderate defects in collapsing the LptFG cavity.

FIG 4.

ΔwaaQ suppresses defects in closing the LptB₂ dimer and LptFG cavity. (A) Table summarizing the phenotype of haploid lptB(E86) variants with and without chromosomal waaQ. OM permeability defects are represented qualitatively with plus signs and are the result of compromised LPS transport. (B) lptB(E86Q) is suppressed by ΔwaaQ. A disc diffusion assay on LB agar was used to assess the OM permeability of haploid mutants to bacitracin (Bac) and rifampicin (Rif) as described in the legend to Fig. 2D. Data are shown as averages from the three independent replicate experiments. Error bars represent the standard deviations. The absence of error bars indicates a standard deviation of zero. NR4446 was used as the parental wild-type lptB strain. (C) ΔwaaQ suppresses lptF(E84A) but not lptG(E88A). Plasmid-borne lptFG alleles were assessed for their ability to complement a chromosomal lptFG deletion both with and without chromosomal waaQ. The OM permeability of lptFG haploid strains was assessed by measuring sensitivity to bacitracin using disc diffusion assays as described in the legend to Fig. 2D. Data show the averages and standard deviations from three independent experiments. NR2761 was used as the parental wild-type lptFG strain. The haploid strain harboring the double glutamate mutant lptF(E84A) lptG(E88A) is unable to complement on LB and is not depicted. Unpaired two-tailed Student’s t test was used to assess the significance of the difference between lptF(E84A) and lptF(E84A) combined with ΔwaaQ. ** indicates a P value of <0.005. (D) Top-down view from the membrane of the Vibrio cholerae LptB₂FGC crystal structure (PDB code 6MJP) shown as a cartoon. The structure is cut away, and only the LptB dimer and the coupling helices (CH) of LptF and LptG are shown. Residues E84 in LptF and E88 in LptG are shown as spheres.

We previously showed that although changing residue E84 in the coupling helix of LptF results in defects in coupling the closure of the LptB dimer with that of the LptFG cavity, changing the equivalent E88 residue in the coupling helix of LptG results in a different, still uncharacterized, defect in LPS transport (36, 37). We found that ΔwaaQ does not suppress defects in OM permeability caused by lptG(E88A), even though, as described above, it suppresses lptF(E84A) (Fig. 4C and D). In agreement, ΔwaaQ changed an lptF(E84A) lptG(E88A) double mutant to resemble the single lptG(E88A) mutant (Fig. 4C). Finally, we also found that ΔwaaQ does not suppress problems in LPS transport caused by defects in either the LptDE OM translocon (Fig. S6A) (53–55) or the initial contacts between LPS and the LptFG cavity (Fig. S6B) (30). From these data, we conclude that ΔwaaQ specifically suppresses defects in the step in the transport cycle when LPS translocates from the LptFG cavity onto the Lpt bridge, which is driven by the simultaneous closure of the LptB dimer and the LptFG cavity (Fig. 1B). Therefore, these data suggest that interactions between the core of LPS and LptB₂FGC affect LPS transport.

lptB(R144H/G236A) is suppressed by changes to three distinct regions of LptFG.

In a previous study, we showed that changes in LptB or the coupling helix of LptF that interfere with the proper closure of the LptFG cavity can be suppressed by changes that either decrease the number of acyl chains on LPS (ΔlpxM suppressor) or alter residues in TM helices of LptG that are located at the bottom of the V-shaped LPS-binding cavity of the LptB₂FGC extractor (37). The acyl chains of LPS reside in this V-shaped cavity prior to extraction, and the TM helices of LptFG must move toward the center of the cavity in order to cause its collapse (31, 33). We therefore proposed that these two types of suppressors promote the movement of LPS from the cavity to the periplasmic bridge, respectively, either by decreasing the number of hydrophobic interactions between the substrate and the cavity or by reducing the energy required to move the LptG helices toward the center of the cavity. Here, our data suggest that the core region of LPS interacts with the transporter in a manner that also affects the extraction process. However, it must act differently than the previously isolated suppressors since the core region does not reside in the hydrophobic LptFG cavity (31, 33).

To identify regions in LptFG that may interact with the core of LPS, we performed selections for suppressors of lptB(R144H/G236A) and targeted those mapping to lptFG (Fig. 5A). We found several suppressors with changes in specific residues in both LptF and LptG that are located in three different regions in the transporter: (i) the coupling helix of LptF [lptF(A90V)], (ii) residues in TM helices of LptFG that align with the bottom of the LPS-binding cavity [lptG(A110V) and lptF(L22F)], and (iii) periplasmic loop 2 in LptF, which links the TM3 helix to the periplasmic β-jellyroll domain [lptF(Δ136–138), lptF(N138T), and lptF(Q148P)] (Fig. 5B). Since the lptG(A110V) allele was previously identified as a suppressor of lptB(E86A), another allele conferring defects in LptB dimer closure, we did not characterize it further (37). However, we included an additional suppressor of lptB(R144H/G236A) that also changes the bottom of the LptFG cavity [lptG(M288R)] and was isolated in a parallel study using an analogous suppressor selection conducted on the lptB1(R144H) double mutant, which encodes the R144H change and a CTD extension (36).

FIG 5.

lptB(R144H/G236A) is suppressed by changes to three separate regions of LptFG. (A) Workflow used to target the identification of lptFG suppressors of lptB(R144H/G236A). Sup refers to suppressor, Mac refers to MacConkey agar, and Tet refers to tetracycline. (1) Cells carrying lptB(R144H/G236A) and yjgN::tet, which is 83% linked to lptFG, were plated on either LB-bacitracin or MacConkey agar to select for spontaneous suppressors. (2) Suppressor colonies were combined into one culture and used to generate a P1_vir lysate. (3) The lysate was used to transduce an lptB(R144H/G236A) strain, and tetracycline-resistant transductants were selected. (4) To identify colonies that acquired yjgN::tet-linked suppressors, transductants were patched onto the original selection medium (either LB-bacitracin or MacConkey agar). (B) Cartoon representation of the apo-LptB₂FGC crystal structure from Enterobacter cloacae (PDB code 6MJP) showing the location of the changes in LptFG, represented as spheres, that suppress lptB(R144H/G236A). LPS (shown as dark blue sticks) was modeled into the LptFGC cavity by aligning the E. coli LPS-bound apo-LptB₂FG cryo-EM structure (PDB code 6MHU) to the E. cloacae structure and showing only LPS. The residues changed by suppressors are color coded by location in light blue (the coupling helix of LptF), orange (along the bottom of the LptFG cavity), and magenta (the periplasmic loop following TM3 and preceding the β-jellyroll domain of LptF). Changes L74P and A110V were previously identified by Lundstedt et al. (37). (C and D) OM permeability of the lptFG suppressors alone and in combination with lptB(R144H/G236A) to bacitracin (Bac) was assessed via a disc diffusion assay on LB as described in the legend to Fig. 2D. Bars are color coded to correspond to the residue coloring scheme in panel B. The data shown are the averages from three independent experiments. Error bars represent the standard deviations. No error bar indicates a standard deviation value of zero. NR4831 was used as the parental wild-type lptB strain.

When lptFG suppressor alleles were reintroduced into lptB(R144H/G236A), we observed that they brought the OM permeability of lptB(R144H/G236A) to either the wild-type or near-wild-type phenotype (Fig. 5C and D). When we introduced these lptFG alleles into an otherwise wild-type strain, we did not observe a change in OM permeability (Fig. 5C and D). As described above for the waaQ suppressor, we next assessed the ability of these lptFG alleles to suppress the conditional loss-of-function allele lptB(R144H). The lptFG suppressor mutations restored the ability of the lptB(R144H) mutant to grow in LB, and the resulting strains displayed minor or moderate defects in OM permeability (Fig. S7). Therefore, these suppressors promote LPS extraction by fixing a defect originally caused by a problem in the closure of the LptB dimer, which extends into a defect in the ability of LptFG to close and squeeze out LPS.

We previously reported that suppressors that alter either the lipid A structure by removing an acyl chain (ΔlpxM) or TM helices in LptG at the base of the V-shaped substrate-binding cavity are synthetically defective (i.e., disproportionately worse) with the CTD-defective lptB-EHis₈ allele (36, 37). Here, we also saw synthetic lethality between the lptB-EHis₈ allele and lptF(L22F) or lptG(M288R), which alter the bottom of the LptFG cavity (Fig. 5B and Fig. S8A). As described above for ΔwaaQ (Fig. S3), we observed an increase in sensitivity when we combined lptB-EHis₈ with the change in the coupling helix of LptF [lptF(A90V)] and one of the changes in the periplasmic loop of LptF [lptF(Q148P)] but not with lptF(Δ136–137) and lptF(N138T) (Fig. S8). These results suggest that suppressors that change interactions between the hydrophobic lipid A structure and the LptFG cavity, as well as those that change residues at the bottom of the LptFG cavity, fix problems caused by defects in LptB dimer closure in a way that is different from that of suppressors that alter LptF’s coupling helix and periplasmic loop 2.

DISCUSSION

The defining feature of ABC transporters is that they physically couple nucleotide-binding domains to substrate-binding transmembrane domains so that they can utilize the energy derived from ATP binding and hydrolysis to transition through steps in the transport cycle (42). How this coupling occurs through long-range distances is poorly understood. In this study, we set out to better understand how the LptB₂FGC ABC transporter coordinates ATP binding with the simultaneous closure of the LptFG cavity and expulsion of LPS onto the periplasmic bridge by identifying suppressors of an lptB mutant with defects in dimer closure. We propose that the suppressors that we characterized reveal sites on the substrate LPS and in LptFG that contribute to this extraction process.

Our work has revealed that changes located in the coupling helix of LptF, TM helices of LptFG, and periplasmic loop 2 in LptF, as well as the loss of part of the core oligosaccharide structure in LPS, suppress problems in LPS transport caused by defects in the closure of the LptB dimer. We previously proposed that the conserved glutamate (residue E84) in the coupling helix of LptF, but not the equivalent residue in that of LptG, plays a role in inducing the closure of the LptFG cavity (37). Finding the lptF(A90V) suppressor provides further support that the coupling helix of LptF is indeed coordinating the movement caused by the ATP-dependent formation of the LptB dimer with the closure of the LptFG cavity. The different function played by the coupling helix of LptG remains unknown.

Together, this work and our previous genetic study have identified residues in TM helices of LptFG along a plane parallel to the membrane that aligns with the bottom of the V-shaped substrate-binding cavity (Fig. 5B) (37). The interior of the LptFG cavity is hydrophobic and accommodates the fatty acyl chains of the lipid A moiety. It has been proposed that LPS is squeezed out of LptFG as the cavity closes because the TM helices of LptFG collapse inward when the LptB dimer closes upon binding ATP (31, 33). Based on the fact that the suppressing changes are located in various TM helices in LptFG but along the same plane (Fig. 5B), we propose that the movement of the TM helices of LptFG is concerted and triggered by residues located at the bottom of the V-shaped cavity. The coordinated inward movement of the helices toward the center of the cavity would resemble that of the closing of a camera iris and push LPS toward the periplasm. The changes to the TM helices encoded by the suppressor mutations likely alter the rigidity of the helices so that they favor cavity collapse by reducing the energy required for the transition.

Our study has also identified that changes to the core of LPS and periplasmic loop 2 in LptF, which connects TM3 to the β-jellyroll domain, suppress defects in the closure of the LptB dimer (Fig. 6). These two types of changes were unexpected since the affected regions are located outside the LptFG cavity. However, recent studies may provide an explanation for how these changes could suppress defects in LPS extraction from the cavity. First, structural and cross-linking studies have identified the β-jellyroll domain of LptF as the first domain in the Lpt periplasmic bridge through which LPS travels after being expelled from the LptFG cavity (32); the lptFG suppressors change residues at the periplasmic loop that is located at the entrance of this β-jellyroll domain (Fig. 6). In addition, cryo-electron microscopy (cryo-EM) structures show the positioning of LPS within the transporter prior to being extracted from the LptFG cavity (31, 33). Although the core oligosaccharide was only partially resolved and the residues altered by the Δwaa suppressors are not visible, the resolved inner-core components are positioned below periplasmic loop 2 of LptF (Fig. 6A). Given the expected location of the core relative to periplasmic loop 2 of LptF, it is therefore tempting to speculate that the suppressing changes altering these domains could be acting in a similar manner. All of the changes in suppressors that alter this region of LptF, Δ(K136-A137), N138T, and Q148P, eliminate potential sites of interaction by removing residues or reducing the size and polarity of side chains. Likewise, waa suppressors produce LPS lacking components, one of them being a charged phosphate (Fig. 3B). Thus, these suppressors could be functioning by removing putative interactions between the core oligosaccharide and the transporter. As LPS makes its way onto the periplasmic bridge from the LptFG cavity, it must pivot to position the acyl chains of lipid A within the β-jellyroll of LptF. Removing potential interactions between the core of LPS and periplasmic loop 2 of LptF could speed up the ability of the core of LPS to rotate and move onto the bridge. In fact, suppressors changing periplasmic loop 2 in LptF might be altering a putative valve that has been proposed to exist at the bottom of the LptF β-jellyroll domain. In the two LptB₂FGC crystal structures, the bottom of the LptF β-jellyroll appears in two different conformations, one open and one closed (32). Introducing cysteines at the base of the LptF β-jellyroll with cysteines placed close enough to form a disulfide bond between them that would be expected to close the entry to the β-jellyroll domain (Fig. 6B) was also shown to prevent LPS from moving onto the bridge unless reducing agents were present (32). These data led Owens et al. to propose that the base of the β-jellyroll of LptF functions as a valve that is needed to prevent the backflow of LPS into the cavity. In most ABC transporters, substrates diffuse away after their translocation (42), but in the Lpt system, LPS that is loaded onto the periplasmic bridge at the β-jellyroll of LptF could potentially slide back into the LptFG cavity after it reopens for the next round of transport. The putative valve in the β-jellyroll of LptF would prevent this problem from occurring. Given the proximity of the substitutions in these suppressors to the putative valve (Fig. 6B), they might influence the rate of its opening/closing in a way that speeds up the movement of LPS onto the bridge.

FIG 6.

Changes in periplasmic loop 2 of LptF that suppress lptB(R144H/G236A) are located adjacent to the core oligosaccharide of LPS and near the putative valve of the β-jellyroll. (A) Cartoon representation of the cryo-EM structure of LPS-bound, nucleotide-free LptB₂FG from E. coli (PDB code 6MHU). LPS is shown as orange sticks, and LptF residues K136, A137, N138, and Q148 are shown as magenta spheres. (B) Cartoon representation of the crystal structure of the nucleotide-free LptB₂FGC from E. cloacae (PDB code 6MJP). For reference (see Discussion), residues S156 and I234 are shown (cyan spheres) and were replaced with cysteines in the structure reported previously by Owens et al. (32), aiming to close the β-jellyroll with disulfide bonds.

We have also uncovered that although all suppressors reported here and elsewhere (36, 37) fix a problem caused by the defective dimerization of LptB, they behave differently when combined with CTD variants that are defective in the opening of the LptB dimer. Specifically, we observed severe synthetic defects when an LptB variant with a defect in its CTD is combined with changes in residues located at the bottom of the substrate-binding cavity in LptFG or mutations that remove acyl chains of LPS but not with changes altering the coupling helix of LptF or the periplasmic loop in LptF (37). The synthetic defects between ΔlpxM and changes to the CTD of LptB must result from a defect when LPS is still in the LptFG cavity that stems from reducing the number of hydrophobic interactions between LPS and the cavity (37). Perhaps closing the cavity too fast by both removing acyl chains in LPS and speeding up the closure of the LptB dimer is detrimental because it miscoordinates specific changes in the transporter that must occur orderly during transport. This miscoordination might also apply to the mutants with an altered CTD of LptB and LptFG cavity residues. Alternatively, these alterations to the CTD of LptB and LptFG cavity residues might facilitate dimer and cavity closure at the expense of greatly disfavoring their opening so that the double mutant complexes cannot efficiently reset after LPS extraction. Nevertheless, the fact that we do not observe such severe synthetic phenotypes when combining changes to the CTD of LptB with changes that alter the coupling helix or periplasmic loop 2 of LptF, or the core of LPS, suggests that these suppressors act by a different mechanism from those altering interactions between the hydrophobic region of lipid A and the LptFG cavity. Since the groove region of LptB, where residue R144 resides, accommodates the coupling helix of LptF, it is reasonable to propose that the A90V change in LptF’s coupling helix directly fixes the miscoupling of LptF and LptB caused by LptB^R144H, thus not causing miscoordination in the transport cycle. We also propose that the above-mentioned models for how changes to the LPS core and to periplasmic loop 2 of LptF may affect LPS transport also explain the absence of severe synthetic defects when they are combined with CTD alterations in LptB since they imply that those changes affect transport at a step that occurs after the collapse of the cavity.

Although more work is needed to fully understand the mechanistic details of LPS extraction, our work has revealed that coupling between the ATPase and its transmembrane domain partners extends beyond the physical contacts between the groove region of LptB and the coupling helices of LptFG and between lipid A and the substrate-binding cavity. We propose that LPS transport also requires coordinating the function of LptB in the cytoplasm with interactions between the core region of LPS and the entrance to the periplasmic bridge of the transporter.

MATERIALS AND METHODS

Growth conditions.

Strains were grown at 37°C in LB or M63 salts medium supplemented with 0.2% (wt/vol) glucose. Cultures were grown either in liquid with aeration or on 1.5% agar plates. When necessary, growth media were supplemented with ampicillin (125 μg/ml), bacitracin (100 μg/ml), chloramphenicol (20 μg/ml), kanamycin (30 μg/ml), tetracycline (25 μg/ml), 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) (33 μg/ml), or isopropyl-β-d-1-thiogalactopyranoside (IPTG) (0.16 mM).

Strain construction and analysis of functionality.

All strains in this study can be found in Table S1 in the supplemental material. The waaGPQY single-deletion alleles from the Keio collection (56) were introduced into strains via P1_vir transduction, selecting for transductants on medium supplemented with kanamycin. The kanamycin resistance cassette was removed via the temperature-sensitive pCP20 plasmid, which encodes Flp recombinase, as previously described (57). Complementation of ΔlptB was assessed on LB and M63 minimal media. All lptFG suppressor alleles linked to yjgN::tet were introduced into strains using P1_vir transduction and selecting for transductants on tetracycline-supplemented LB plates. The chromosomal lptFG region was amplified by PCR and sequenced to confirm the presence of these alleles. All primers used are listed in Table S2 in the supplemental material. pRC7CatSacBLptB-containing strains were built for each waa and lptFG suppressor to assess complementation of chromosomal ΔlptB by plasmid-borne lptB alleles (29).

Antibiotic sensitivity testing.

Assessment of OM permeability was accomplished by testing the cells for antibiotic sensitivity using a disc diffusion assay (29). After growing cells overnight in liquid LB, 50 μl of the culture was mixed with 4 ml of LB top agar (0.75% agar) and poured on LB agar plates (1.5% agar). If strains had growth defects with respect to the wild-type strain, the amount of cells used was normalized by the optical density at 600 nm (OD₆₀₀) prior to mixing with LB top agar. Discs containing bacitracin, novobiocin, erythromycin, and rifampicin were placed on top of the agar once it solidified. Following overnight incubation at 37°C, zones of full and partial inhibition of growth around the discs were measured.

Suppressor selection and mapping.

The waaP::IS1 suppressor was selected by growing a culture of NR4447 [haploid lptB(R144H/G236A)] overnight. A 100-μl sample of this culture was plated on LB agar plates containing bacitracin and incubated overnight. Colonies that grew contained a suppressor. The waaP::IS1 allele was mapped by P1_vir cotransduction frequency to tetracycline-resistant markers as described previously (58). The suppressor mutation was identified by amplifying the chromosomal waaP locus via PCR and sequencing the resulting PCR product.

For the targeted selection of lptFG suppressors, we used NR4550, a haploid lptB(R144H/G236A) strain harboring yjgN::tet, a mini-Tntet transposon insertion that has an 83% linkage to the lptFG locus. NR4550 was grown overnight in LB, and 100 μl of the culture grown overnight was plated on either MacConkey or LB-bacitracin (100 μg/ml) agar. Colonies that grew under these conditions had acquired a suppressor and were combined to generate a sup⁺ tet⁺ P1_vir lysate. A culture of NR4777 [haploid lptB(R144H/G236A)] was grown overnight in LB and transduced with the sup⁺ tet⁺ lysate. Transductants were selected on LB-tetracycline agar overnight. Colonies were patched onto their original selection medium (MacConkey agar or bacitracin) to screen for sup⁺ tet⁺ strains. The chromosomal lptFG locus was amplified via PCR and sequenced to identify suppressors.

Immunoblotting.

Cells were grown overnight in LB, and whole-cell lysates were prepared and normalized by the cell density as previously described (29). The samples were subjected to 10% (for assessing LptB levels) or 15% (for assessing LPS levels) SDS-polyacrylamide gel electrophoresis and subsequently transferred onto polyvinylidene difluoride membranes (29). Samples were probed with rabbit anti-LptB (1:10,000 dilution) or mouse anti-LPS (1:10,000 dilution) (catalog number 4329-5004; Bio-Rad) primary antibodies. Following primary antibody exposure, samples were exposed to anti-mouse horseradish peroxidase-conjugated secondary antibodies (1:10,000) (GE Healthcare Life Sciences) or anti-mouse antibodies (1:10,000) (GE Healthcare Life Sciences), respectively. The signal was developed using the Clarity Western ECL substrate (Bio-Rad), imaged with a Chemidoc CRS⁺ system, and visualized with ImageLab 5.2.1 software (Bio-Rad).